Dear @gastonstat,
As far as I was able to check, neither plspm nor any of the other PLS SEM packages currently available in R is able to perform Weighted PLS taking into account the already available weights from a complex survey that used stratified sampling. A few weeks ago I was introduced to two articles (https://doi.org/10.1016/j.emj.2016.06.009 and https://doi.org/10.1080/14783363.2020.1754125) that showed the importance of using WPLS to analyze data from such surveys.
The WPLS algorithm that was used in both articles was implemented in SmartPLS. Currently, there is the lavaan.survey package in R which adjusts the results of lavaan objects taking into account the survey weights. I believe that if it is possible to implement the WPLS algorithm in plspm, this would markedly increase the possibilities of using PLS in fields such as epidemiology and medicine.
If my understanding was incorrect and there is some way of adjusting the results of plspm models for sampling weights, I would be grateful if you could provide some guidance in that regard.
On the other hand, if at this moment plspm is not compatible with WPLS and you would consider adding this functionality to it and you need any further details regarding examples of datasets with complex sampling designs or on the use of the survey package in R, which is commonly used to analyze such data, please do not hesitate do contact me at edison.vidal@unesp.br.
Best wishes,
Edison
Dear @gastonstat,
As far as I was able to check, neither plspm nor any of the other PLS SEM packages currently available in R is able to perform Weighted PLS taking into account the already available weights from a complex survey that used stratified sampling. A few weeks ago I was introduced to two articles (https://doi.org/10.1016/j.emj.2016.06.009 and https://doi.org/10.1080/14783363.2020.1754125) that showed the importance of using WPLS to analyze data from such surveys.
The WPLS algorithm that was used in both articles was implemented in SmartPLS. Currently, there is the lavaan.survey package in R which adjusts the results of lavaan objects taking into account the survey weights. I believe that if it is possible to implement the WPLS algorithm in plspm, this would markedly increase the possibilities of using PLS in fields such as epidemiology and medicine.
If my understanding was incorrect and there is some way of adjusting the results of plspm models for sampling weights, I would be grateful if you could provide some guidance in that regard.
On the other hand, if at this moment plspm is not compatible with WPLS and you would consider adding this functionality to it and you need any further details regarding examples of datasets with complex sampling designs or on the use of the survey package in R, which is commonly used to analyze such data, please do not hesitate do contact me at edison.vidal@unesp.br.
Best wishes,
Edison